teleost fish scales amongst the toughest collagenous...

Available online at www.sciencedirect.com

www.elsevier.com/locate/jmbbm

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 5 2 ( 2 0 1 5 ) 9 5 – 1 0 7

http://dx.doi.org/101751-6161/& 2018 El

nCorresponding aE-mail address: f

Research Paper

Teleost fish scales amongst the toughestcollagenous materials

A. Khayer Dastjerdi, F. Barthelatn

Department of Mechanical Engineering, McGill University, Montreal, QC, Canada

a r t i c l e i n f o

Article history:

Received 27 May 2014

Received in revised form

23 September 2014

Accepted 27 September 2014

Available online 6 October 2014

Keywords:

Fish scales

Collagen

Toughness

Process zone

Crack bridging

Bio-inspired materials

.1016/j.jmbbm.2014.09.025sevier Ltd. All rights rese

uthor.rancois.barthelat@mcgill

a b s t r a c t

Fish scales from modern teleost fish are high-performance materials made of cross-plies of

collagen type I fibrils reinforced with hydroxyapatite. Recent studies on this material have

demonstrated the remarkable performance of this material in tension and against sharp

puncture. Although it is known that teleost fish scales are extremely tough, actual

measurements of fracture toughness have so far not been reported because it is simply

not possible to propagate a crack in this material using standard fracture testing

configurations. Here we present a new fracture test setup where the scale is clamped

between two pairs of miniature steel plates. The plates transmit the load uniformly,

prevent warping of the scale and ensure a controlled crack propagation. We report a

toughness of 15 to 18 kJ m�2 (depending on the direction of crack propagation), which

confirms teleost fish scales as one of the toughest biological material known. We also

tested the individual bony layers, which we found was about four times less tough than the

collagen layer because of its higher mineralization. The mechanical response of the scales

also depends on the cohesion between fibrils and plies. Delamination tests show that the

interface between the collagen fibrils is three orders of magnitude weaker than the scale,

which explains the massive delamination and defibrillation observed experimentally.

Finally, simple fracture mechanics models showed that process zone toughening is the

principal source of toughening for the scales, followed by bridging by delaminated fibrils.

These findings can guide the design of cross-ply composites and engineering textiles for

high-end applications. This study also hints on the fracture mechanics and performance of

collagenous materials with similar microstructures: fish skin, lamellar bone or tendons.

& 2018 Elsevier Ltd. All rights reserved.

1. Introduction

Over millions of years of evolution, biological organisms havedeveloped high-performance natural materials to fulfill avariety of structural functions. Materials like bone, teeth,scales or mollusk shells possess outstanding mechanicalproperties despite their relatively weak ingredients. During

rved.

.ca (F. Barthelat).

the past decade much research has been devoted to under-standing the concepts, structures and mechanisms under-lying the performance of these natural materials, so they canbe implemented in high performance synthetic materials(Barthelat, 2007; Bonderer et al., 2008; Mayer, 2005; Studart,2012; Mirkhalaf et al., 2014). Mollusk shells, arthropod cuti-cles, bone, teeth have already attracted a great deal of

http://dx.doi.org/10.1016/j.jmbbm.2014.09.025




http://crossmark.crossref.org/dialog/?doi=10.1016/j.jmbbm.2014.09.025&domain=pdf

mailto:[email protected]



attention, and more recently many more natural materialshave emerged as potential source of inspiration for newbiomimetic materials (Meyers et al., 2008; Dimas andBuehler, 2012; Weaver et al., 2012). The structure andmechanics of natural scaled skin, and more particularly fishscales, have recently been the subject of several studies (Yanget al., 2013a,b; Bruet et al., 2008; Ikoma et al., 2003; Garranoet al., 2012; Lin et al., 2011; Meyers et al., 2012; Zhu et al., 2012,2013; Browning et al., 2013; Zimmermann et al., 2013;Vernerey and Barthelat, 2010). Since their emergence 500million year ago, fish have evolved scales with differentshapes, sizes and arrangements which are generally classi-fied into cosmoid, ganoid, placoid and elasmoid (Benton,2004). Over the course of evolution the bony multilayeredcosmoid and ganoid categories have largely been replaced bythinner yet more flexible teleost scales (Kardong, 2006).Natural scaled skins have remarkable mechanical properties:compliance, resistance to penetration, lightweight, and ultra-thin structure (Yang et al., 2013a,b). Several studies on thestructure and mechanics of fish scales were performed sinceon ancient bony scales (Bruet et al., 2008; Yang et al., 2013a,b)and on the more modern and lighter teleost scales (Ikomaet al., 2003; Garrano et al., 2012; Lin et al., 2011; Zhu et al.,2012). Mechanical characterization typically involved tensiletesting (Ikoma et al., 2003; Garrano et al., 2012; Lin et al., 2011;Zhu et al., 2012) or indentation and puncture tests onindividual scales (Bruet et al., 2008; Meyers et al., 2012; Zhuet al., 2012) or multiple scales (Zhu et al., 2013). Fish scaleshave also started to inspire new flexible protective systems(Browning et al., 2013; Chintapalli et al., 2014). Tensile tests onnatural teleost fish scales confirmed the scale as a stiff,strong a tough material. In collagenous scales extensiveinelastic deformation and energy dissipation were observedincluding pullout, defibrillation, sliding and rotation (Ikomaet al., 2003; Garrano et al., 2012; Lin et al., 2011; Zhu et al.,2012; Zimmermann et al., 2013). The function of the scaledskin is to provide protection while maintaining high compli-ance in flexion to allow for motion (Vernerey and Barthelat,2010). One of the main functions of the scale is thereforemechanical protection against predators, collisions withother fish or obstacles or other mechanical threats. Recentstudies have therefore focused on the indentation and punc-ture resistance of individual scales (Bruet et al., 2008; Zhuet al., 2012), showing that scales are superior to engineeringmaterials such as polycarbonate and polystyrene (Zhu et al.,2012). This high performance can be explained by the abilityof the material to dissipate large amounts of energy throughlarge deformations. In turn, these large deformations arepossible because the material can resist the propagation ofcracks that could emanate from pre-existing defects in thematerials and/or from high stress concentrations typical topuncture or laceration. In this context high fracture tough-ness becomes a desirable property, and early tests noted thatindeed some fish scales are so tough that they could not befractured, even after immersion in liquid nitrogen (Currey,1999). Fracture tests on nonlinear, extensible and toughmaterials are in general difficult using traditional fracturetesting configurations. For elastomers the “trouser” tear testcircumvents some of the experimental difficulties (ASTM,2003), and it is successfully applied on biological elastomers

such as skin (Purslow, 1983). However, this test provides amode III fracture toughness, which may differ from mode Ifracture especially if fracture process is governed by defibril-lation and fiber bridging as in fish scales. Instrumented modeI fracture tests on individual fish scales are extremely difficultbecause of their small size and high fracture toughness. Tothe best of our knowledge, toughness could only be measuredon scales from gar, which are heavy ganoid scales withsufficient size, thickness and stiffness to permit toughnessmeasurement in bending (Yang et al., 2013a,b). Gar scaleswere found similar to cortical bone in terms of composition,structure and fracture toughness (3 to 5 MPa m1/2 (Yang et al.,2013a,b)). However it has so far been impossible to measurethe toughness of teleost fish scales because there are thinner,smaller, and because they undergo massive inelastic defor-mation. The toughness of this type of scale appears to be sohigh that it is not possible to fracture them using conven-tional fracture test configurations. In this paper we demon-strate that teleost fish scales from striped bass (Moronesaxatilis) are notch insensitive, an indication of high tough-ness. Using a new miniaturized fracture test setup we thenreport, for the first time, the fracture toughness of teleostscales from striped bass (M. saxatilis) along different direc-tions. The fracture toughness of the bony layer, collagen layerand non-collagenous interfaces were also measured. Wefinally use a set of simple fracture mechanics models toassess the main contributors of the high toughness of fishscales.

2. The hierarchical structure of scalesfrom striped bass Morone saxatilis

Like many other biological materials (Fratzl and Weinkamer,2007), fish scales display a hierarchical structure organizedover several length scales (Fig. 1) (Zhu et al., 2012). Individualscales of striped bass are roughly pentagonal in shape andabout 8–10 mm in diameter. Topological features on thesurface of the scale, together with the underlying microstruc-ture define different sectors on the scale: anterior (A), poster-ior (P) and lateral (L) regions. The anterior region of scalesexhibits a triangular shape with radial grooves (radii) andcircular rings (circuli) forming around the central area of thescales called “focus” (Zhu et al., 2012). While radii are onlypresent in the anterior field, circuli also appear in the poster-ior region but their circular form is not well recognized (Zhuet al., 2012). It is believed that radii promote the flexibility ofthe scales and circuli are involved in the anchoring of thescale into the dermis (Zhu et al., 2012). Scales are the thickestat the focus region (300–400 μm), and the thickness decreasescontinuously towards the edges. A cross section of the scalereveals two distinct layers: the outer layer which is partiallymineralized (hydroxyapatite volume fraction of 30% (Zhuet al., 2012)) and often called “bony layer”, and the innerlayer which undergoes less mineralization (hydroxyapatitevolume fraction of 6%) and is referred to as “collagen layer”(Fig. 1). Each layer is composed of approximately 10–20 plieseach �5 μm thick, which are composed of fibrils of collagentype I. Plies “R” are composed of fibrils aligned along theradial direction while plies “C” are composed of fibrils

Bon

y la

yer

Col

lage

n la

yer

10-3 10-610-4 10-5 meters

2mm 50μm 5μm 500nm

Ant

erio

r (A)

Post

erio

r (P

)

Focus

Fig. 1 – The hierarchical structure of individual fish scales, from mesoscale to nanoscale. (adapted from (Zhu et al., 2012)).

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 5 2 ( 2 0 1 5 ) 9 5 – 1 0 7 97

forming concentric pentagons centered on the focus of thescale. Plies “R” and “C” are alternatively stacked to form amultilayered structure. At the local scale, the collagen fromthe “R” and “C” layers cross at a right angle, locally forming across ply microstructure (Zhu et al., 2012). A compound ofnon-collagenous proteins hold the collagen fibrils together.These ingredients are found in virtually all collagen-basedmineralized tissues such as bone, with the ratio of thesecomponents varying with the function of the tissue (Currey,1999). Interestingly, unlike collagen in skin, tendon or liga-ments, the collagen fibrils in fish scales are relatively straightand do not display the crimps which are often present inother soft collagenous tissues (Zhu et al., 2012).

3. Notch performance

The ability of the material to redistribute stresses aroundcracks or imperfections to eliminate stress concentration anddelay fracture is critical to the robustness of both engineeringand biological materials (Evans, 1997; Evans et al., 2001). Hightoughness and inelastic deformations are required for aredistribution of stresses to operate, and studies on engineer-ing materials have shown that materials which can undergotensile inelastic strain of at least four times the elastic straincompletely suppress stress concentration notch insensitive(Davis et al., 2000). Fish scales are subjected to extremeloading which are concentrated over small surfaces andvolumes in the case of a predator’s attacks, and resistinghigh stress concentrations is critical to the performance ofthe scale and to the survival of the fish. Previous studieson the tensile response of fish scales from striped bassshowed maximum elastic strains of about 5% and strain atfailure of about 30%, accompanied with extensive inelasticdeformations. These large deformations suggest that fishscales are notch insensitive, and in this section we actually

tested this hypothesis. Fresh striped bass (M. saxatilis) waspurchased from a fish supplier (Nature’s Catch, Inc., Clarks-dale, MS, USA), and scales were plucked using tweezers andstored in a freezer at �20 1C until testing. Prior to testing thescales were thawed in water at room temperature for 5 min.In order to assess the notch sensitivity of fish scales weprepared two batches of samples: the first batch consisted oftensile specimen prepared by cutting away two half-disksfrom the lateral regions of individual scales using a sharpmulti-tube rotary hole-punch (5.6 mm diameter). This simpleprocess produced a dog-bone shape tensile specimen with agage width of about 4 mm (Fig. 2a). The second batchconsisted of the same sample geometry, with the additionof a deep notch introduced with a razor blade about half-wayacross the width of the sample (Fig. 2b). The prepared intactand notched tensile specimens were then mounted on aminiature loading stage (Ernest F. Fullam Inc., Latham, NY) tobe tested along the anterior–posterior direction. All the testswere performed in hydrated conditions, while a digitalcamera was used to record pictures during the course ofthe tests.

Fig. 2c shows typical force–displacement curves for intactand notched specimens. The curves have a bell shape, withan initial linear response followed by large tensile strains andtremendous energy absorption. As expected, the notchedsamples were weaker than the intact samples because ofthe presence of the notch and the associated reduction in thenominal cross section (i.e. minimum load bearing crosssection). Based on this data, the nominal stress was com-puted by dividing the tensile force by the nominal crosssection. The cross section used to compute the stresses in theintact sample was obtained by multiplying the width ofthe sample in the middle of the gage region (�4 mm) bythe sample thickness (�300 μm). For the notched sample thenominal cross section is the width of the ligament after thenotch was introduced (�2 mm) multiplied by the sample

Notch

Bony layer fractures

4 mm

Notched

Notched

Intact

Intact

Fig. 2 – Schematic illustration of (a) intact tensile sample and (b) notched tensile sample; (c) typical force–displacementfor intact and notched samples and (d) corresponding nominal stress–displacement curves.


thickness (�300 μm). The exact dimensions of the crosssection were measured for each sample. Fig. 2d displaystypical nominal stress–displacement curves obtained forintact and notched tensile specimens. Note that these experi-ments were intended to compare nominal stresses in notchesand intact samples, and the exact strains in the sample werenot computed (full tensile stress–strain curves for this type offish scale can be found elsewhere (Zhu et al., 2012)). Remark-ably the curves for the intact and notched samples arevirtually identical. We found a nominal strength of32.276.6 MPa for the intact samples and a nominal strengthof 35.074.2 MPa for fracture samples (N¼5 samples tested foreach group). The failure modes in intact and notched sampleswere also identical: at the early stage of loading, the bony andcollagen layer delaminated because of mismatch betweentheir mechanical properties (Zhu et al., 2012) (Fig. 3b). Furtherincrease of loading resulted in the fracture of bony layerwhile the collagen layer was still deforming, with extensivedefibrillation of the collagen cross plies up to the ultimatefailure at around 0.8 mm displacement. Because of variationin the thickness of the bony layer (bony layer is thicker atfocus region), the fracture of bony layer in all the samplesoccurred at a region between the clamped site and notchedarea. Intact and notched samples also displayed stresswhitening, an indication of delamination between bony andcollagen layer occurs and also of region at which damage isaccumulated (mainly in the bony layer). In bony and collage-nous materials, whitening has also been associated in thepast with inelastic deformation and with the formation ofligaments (Thurner et al., 2007). The notched sample alsoshowed crack blunting (Fig. 3b and c), a potent toughening

mechanism for metals and polymeric materials. Blunting wasaccompanied by a whitening of the material ahead of thenotch and eventually across the entire ligament.

Finally the ligament ahead of the notch completely defi-brillated and collapsed in tension, with no apparent crackpropagation. This mechanism is similar to tendon, anotherhigh-performance material composed of unidirectional col-lagen fibrils (Ker, 2007). Fish scale are therefore notch insen-sitive: the stress concentration introduced by the notch iscompletely suppressed by micro-mechanisms associatedwith the cross-ply structure of the scale, and the onlydecrease in apparent strength is due to the reduction innominal cross section. The toughening mechanisms are sopowerful in fish scales that the pre-crack simply does notpropagate, while the ligament undergoes massive non-lineardeformations and yielding. Since the sample fails by collapseof the ligament instead of fracture, fracture toughness mea-surements are not possible from this configuration. Wetherefore developed another approach to force crack propa-gation through the scale, which we describe in the nextsection.

4. Fracture toughness of the whole scale

The experiments in the previous section showed that fishscales from striped bass are notch insensitive even in thepresence of a deep notch half way across the scale. Thesamples failed by inelastic deformation and collapse ratherthan crack propagation. Therefore stress concentrations,generated for example by a sharp indenter, can be ignored

Initial notch

Bony layer fractures

Fig. 3 – Sequence of images taken during the notch performance test with diagrams: (a) initial geometry of the notchedsample; (b) bony layer fractures and delaminates; (c) crack blunting and propagation of inelastic region in collagen layer;(d) extensive defibrillation of the collagen layer and massive inelastic deformations in the collagen layer. There is no obviouscrack propagation. The red region indicates inelastic deformation and processes.


in the scales and a criterion based on “strength” or “yield-ing” would be appropriate and sufficient to predict thefailure of these scales. Nevertheless, it is useful to deter-mine fracture toughness in the more general context of thefracture mechanics of collagenous tissue and of bio-inspiration. Notch insensitivity is an indication of hightoughness, so in terms of designing notch insensitivematerials it is useful to know what the toughness of thescales is. Toughness can also be correlated with structureand micromechanics to provide more insights into materialdesign. The toughness of scales can also be compared withthe toughness of existing collagenous tissues, with usefulinsights into structure–property relationships. In the nat-ural fish scale we tested, the notch length and sample size isin the same order as the process zone. In a hypotheticallarger fish scale, or a possible biomimetic fabric inspiredfrom fish scales, the size of the notch and of the componentcan be much larger, and material nonlinearities may not besufficient to suppress the effects of stress concentration atthe crack tip (as seen in oxide fiber composites (Zok, 2006)).In these cases, failure may be governed by fracturemechanics rather than inelastic deformations and “yield-ing”. Measurements of toughness on the fish scales fromstriped bass proved challenging. The scales are relativelycompliant, small and extremely tough, and all the conven-tional fracture tests configurations we tried were unsuc-cessful in propagating cracks and in producing meaningfulmeasures of toughness. Compact tension fracture tests(with pre-notch and forces transmitted though pins)resulted in large out-of-plane warping of the samples.In bending configurations, the sample underwent large deflec-tions without any crack propagation. In order to overcome these

limitations we used a pair of miniature custom made miniaturefixtures made of stainless steel in order to clamp the scales(Fig. 4). This configuration promotes crack propagation and limitsin-plane and out-of-plane deformations. Slippage at the clamp isalso limited, because large surfaces of the scale are used forclamping. The test geometry is similar to the rigid doublecantilever beam (RDCB) fracture test, which we recently devel-oped to measure the fracture toughness of biological proteinsand engineering adhesives (Dastjerdi et al., 2012). Here thefracture samples were prepared by punching six holes (1mmin diameter) though the scale: two pairs of holes were used asscrew holes to clamp the scale with the two pairs of rigid steelfixtures, while the other pair was used as pin holes to exertopening forces with the miniature loading machine (Fig. 4aand c). In order to monitor the crack propagation process, agap of �1mm was kept between the two clamps, so the sampleconfiguration consisted of a rectangular strip of fish scaleclamped to two rigid plates. After clamping the scale, a deepcut about 3mm long was introduced on the side of the scalewith a fresh razor blade. The clamped samples were finallyplaced in a miniature loading stage mounted with two pins, andinstalled under an upright, reflected light microscope (BX-51M,Olympus, Markham, Canada) equipped with a CCD camera(RETIGA 2000R, Qimaging, Surrey, Canada) to monitor crackpropagation. All tests were performed at an opening displace-ment rate of 5 μms�1 and in a fully hydrated condition (the gapbetween the fixtures was filled with water). In order to investi-gate possible anisotropy in toughness, cracks were propagatedalong three different directions: anterior–posterior (A–P), lateral–lateral (L–L) and posterior–anterior (P–A) (Fig. 4b).

Crack propagation was very stable, an indication of power-ful toughening mechanism within the scale. The force–

Fig. 4 – (a) Preparation and testing of the fracture sample; (b) the three orientations explored for crack propagation; (c) actualpicture of an ongoing fracture test.


displacement curves (Fig. 5a) also displayed a bell shape, acharacteristic of tough materials. All curves showed an initialquasi-linear region with a stiffness of 15–20 Nmm�1, followedby softening as damage accumulated and the crack propa-gated in the fish scale. Damage accumulation continued untila maximum load, after which the load progressivelydecreased. Examinations of in-situ optical images revealedthat the sudden drop of force after the point of maximum loadcorresponds to the brittle fracture of the bony layer. Fig. 5bshows an optical image captured just after the peak load forA–P crack propagation, showing an entirely fractured bonylayer. At this point the underlying collagen has undergonemassive defibrillation, but intact fibrils are still bridging theopening crack. The water saturating the crack made imagingdifficult, but defibrillated collagen at 30–601 from the crackfaces could be observed behind the crack tip (Fig. 5c). Fig. 5dshows a schematic of the structure of these fibrils. The fibersbelonging to the circumferential layer “C” crossed the propa-gating crack at 901. These fibers were pulled out from thecollagen layer, and probably failed at moderate crack openings.Another type of fibers from the radial “R” layer delaminatedfrom the collagen layer as the crack propagated, continuouslyfeeding collagen fibrils into the opening crack. We observedmore of these fibers than the “C” fibers, probably because thefibers from the “R” layer could bridge the crack over greatercrack opening distances. In summary, once the brittle bonylayer has fractured, simultaneous crack bridging of fibers fromthe “R” and “C” layer appears to be the main cohesivemechanism in fish scales. Crack bridging is a powerful tough-ening mechanism which was observed in other natural com-posites including bone (Nalla et al., 2003), nacre (Zhu et al.,2013), and tooth enamel (Imbeni et al., 2005).

Here we determined the fracture toughness of the scalesusing the “work of fracture” which is the energy (area under

force–displacement curve) required to fracture the notchedspecimen divided by the area of the fractured surface. Workof fracture can also be interpreted as the average energy perunit thickness required to propagate the crack by a unitdistance, and in that sense it directly correlates with theenergy definition of fracture toughness based on J-integral.Fig. 5b shows the work of fracture obtained for the threedifferent directions of crack propagation: anterior–posterior(A–P), lateral–lateral (L–L), and posterior–anterior (P–A). Wefound a toughness of 17.670.5 kJ m�2, 15.371.1 kJ m�2, and18.470.7 kJ m�2 for A–P, L–L, and P–A directions, respectively,which are remarkably high values compared to bone(1–10 kJ m�2) (Koester et al., 2008) or even mammalian skin(10–20 kJ m�2) (Purslow, 1983). The fracture toughnessappears to be slightly lower in the L–L direction, and furtherstatistical analysis (Student’s t-test) confirmed that the workof fracture along the L–L direction is statistically significantlylower (po0.05) than in the A–P and P–A direction, which werestatistically the same. A slightly lower toughness in the L–Ldirection can promote crack propagation across the scale,which may ultimately result in “clipping” of the scale andminimal consequences for the animal. In contrast, crackspropagating towards the anterior region of the scales may bemore dangerous, because they can provide pathways tobacteria and other external agents towards the softer under-lying tissues. We conclude this section by a few remarks onthe failure model and measurement of work of fracture. Sincethe scale is notch insensitive for this specimen size and notchlength, the failure was dominated by inelastic collapse. Asample with a shorter notch or even no notch would producethe same work of fracture. In either of these cases, a mode Itearing of the scale is imposed by the progressive opening ofthe stiff clamps. In practice, testing specimens with shorternotches was not possible because the samples slipped out of

A-P

whole scale

collagen layer

bonylayer

L-L P-A L-L L-L

Col

lage

n fib

rils

A-P

L-L

P-A

Fig. 5 – (a) Typical force–displacement curves obtained from fracture experiment for the three directions of crack propagation:(b) corresponding work of fracture; (c) optical image taken during the softening region of the force–displacement curve(A–P crack propagation) showing the fractured bony layer and extensive defibrillation of the collagen layer; (d) schematicshowing the structure of the defibrillated collagen during the course of crack opening.


the grips before crack propagation started. During the test, wealso could not observe and locate a crack tip in the traditionalsense, because of blunting, defibrillation and massive inelas-tic deformations. It is in fact likely that the inelastic regionspreads across the entire scale ahead of the notch before orshortly after crack propagation occurs. Nevertheless, theresults from the experimental setup used here directly reflectthe traction–separation behavior of the material, and we haveindeed recently used a similar setup to determine cohesivelaws of various adhesive in the past, with modeling andvalidation against independent toughness measurements(Dastjerdi et al., 2012, 2013). In this experimental configura-tion, the energy dissipated up to failure can therefore be usedto compute work of fracture and toughness, even if theinelastic region extends along the entire sample.

5. Fracture toughness of bony and collagenlayers

Crack propagation across the scale involves fracturing bothbony and collagenous layers, two materials with distinctcompositions and high contrast of properties. In the contextof developing biomimetic systems inspired from scales, it isimportant to determine the individual properties for the bony

and collagenous layers. In this section we measured thefracture toughness of each of the individual layers, usingdirect measurements and indirect methods. Following proto-cols developed in previous work on fish scales (Zhu et al.,2012), the collagen layer was carefully ablated from a scale.The remaining material, made entirely of bony material, wasmore brittle than the collagen layer, so that traditionalfracture toughness tests could be used. The bony layer wascut into a dog-bone shape (with gage width and length of2 mm) using a laser engraver (Mirkhalaf et al., 2014) with itslongitudinal direction parallel to the A–P axis. A 1.5 mm longpre-notch was then introduced across the gage region usinglaser engraving. During the whole process of cutting, thescales were kept hydrated in order to cool the sample and toavoid shrinkage of the samples from drying. The sampleswere then tested in tension using the miniature loading stageand hydrated conditions. The samples showed much lessdeformations compared to the whole scale, but fracturepropagation was still stable. We found a work of fracture ofJB¼7.370.7 kJ m�2 for the bony layer, which is about twotimes lower than the work of fracture for the whole scale.Meanwhile, none of the samples of pure collagen obtainedfrom the ablation protocol could be tested for fracturetoughness as they would invariably slip out of the grips.The toughness of the collagen layer could however be


estimated from the toughness of the whole scale and of thebony layer. Propagating a crack through the whole scalesinvolves the fracture of both bony and collagen layers. Sincethese two layers have approximately the same thickness, thefracture toughness of the whole scale JS can be estimated by:

JS �12

JC þ JBð Þ ð1Þ

where JC and JB denote the fracture toughness of the collagenand bony layer, respectively. Eq. (1) does not take in accountthe interactions between bony and collagen layers (whichprobably exist), but provides a means to evaluate the tough-ness of the collagen layer which we could not obtain experi-mentally. JS and JB are known from the experiments describedabove, and Eq. (1) yielded a fracture toughness of JC¼23.572.9 kJ m�2 for the collagen layer, which is more than threetimes the toughness of the bony layer. While both bony andcollagen layers are composed of collagen fibrils with almost thesame arrangement, it is probably the higher concentration ofbrittle hydroxyapatite within the bony layer which results in itslower fracture toughness compared to the collagen layer.

6. Interlaminar fracture toughness withinindividual scales

Collagen is a high-performance fiber, and its high deform-ability and high strength generate, for a large part, thetoughness of fish scales. Equally important is the cohesionbetween the collagen fibrils. As observed above pullout,delamination and defibrillation play an important role inthe toughening mechanisms operating in the scale, and thesemechanisms are governed by the toughness of the extra-collagenous proteins at the boundaries between the fibrils. Inorder to determine the interlaminar fracture toughness ofindividual scales, we performed a series of peel tests inhydrated condition. For this test, the scales were cut intorectangular strips of 8�4 mm, with their length parallel tothe A–P axis of the scales (Fig. 6a).

In order to facilitate handling and testing, the scale wasdemineralized in Surgipath Decalcifier II (Leica MicrosystemsInc., Buffalo Grove, IL) following standard procedures. Usinga fresh razor blade the scale was partially delaminated inorder to introduce a pre-notch about 4 mm deep. The ends of

Clamps

Demineralized bony layer

Collagen layer 4 m

Anterior

4 mm

Fig. 6 – (a) Picture showing the peel test sample geometry and omeasuring the delamination toughness of individual scales; (c) tw

the sample were then clamped in the miniature loading stageas shown in Fig. 6b. The samples were then loaded at a rate of5 μm s�1 until the material was completely debonded. Thesamples were kept hydrated throughout the tests by frequentirrigation with water. The propagation of the interlaminarcrack was very stable until the end of the test where thecollagen and bony layer completely separated, and post-mortem imaging revealed that the crack tended to followthe interface between bony and collagen layer. Fig. 6c showstwo typical force–displacement curves from this test. Follow-ing elastic loading, delamination of the scale commenced at aconstant force of about 0.05 N. At a displacement of about5 mm (corresponding to a crack propagation of 2.5 mm), theforce increased continuously up to about 0.12 N at which thestrip completely delaminated. The ultimate force wasreached at a displacement of 8 mm, corresponding to a crackadvance of 4 mm (which was the length of the remainingmaterial after the initial notch was produced. The increase offorce from 5 to 8 mm (corresponding to the anterior and focusfield of individual scales) indicates that the interlaminarstrength of collagen layers most likely increases from ante-rior to focus area. The toughness of the interface, in energyterms, was simply obtained by dividing the area under theforce–displacement curves by the area of the fractured sur-face. This calculation yielded a toughness of 38.679.8 J m�2.For comparison this delamination toughness is about twentytimes higher than office tape on a glass substrate (Dastjerdiet al., 2012), and more than ten times higher than thetoughness of the interfaces in nacre (Dastjerdi et al., 2013).The delamination toughness is 400 times lower than thetoughness of the whole scale measured above which isbeneficial, considering that low interfacial toughness is arequirement for pullout, delamination and defibrillation ofthe collagen in the fracture process of the whole scale in theA–P, L–L and P–A directions.

7. The fracture mechanics of fish scales

The previous sections demonstrated the high toughness offish scales from striped bass, suggesting that this materialcould serve as model for bio-inspired protective materialsand systems. However, in order to establish design principles

m

Focus

300 μm

rientation; (b) schematic illustration of peel test setup foro typical force–displacement curves obtained from this test.

2w

Unloaded materialInelastic region


for such material, a deeper understanding of the fracturemechanics of fish scales is required. In this section wepresent models which capture the crack resistance producedby collagen fibrils. The first model examines the contributionof process zone toughening, where the energy dissipated in astrip of material around the crack generates fracture tough-ness. In a second section we examine the potential contribu-tion of fiber delamination on crack bridging. The modelsfocus on the fracture mechanics of the tougher collagen layer.Similar mechanisms may operate in the bony layer but to alower extent, because the mineralization of the fibrils limitstheir extension.

1

Fig. 7 – Process zone toughening in the collagenous layer ofthe fish scale in a steady state configuration. The frontalzone (size 2w) leaves a wake of inelastically deformedmaterial behind the crack tip.

7.1. Process zone toughening

Experimental observations showed that massive nonlineardeformations ahead of the crack and in its wake are aprominent feature in the process of fracturing the scale.Fig. 7 schematically shows a crack propagating through thecollagenous layer, where the material “yields” just ahead ofthe crack tip because of high stresses. Once the crackpropagates through the initial inelastic region the materialsnear the crack face unloads, and the process of loading/unloading dissipates energy (Shen et al., 2008). Assuming thatthe yielding is governed by tensile stress only, the half widthof the yielded region is given by Evans et al. (1986), Barthelatand Rabiei (2011):

w¼ 0:25JCEC

σC2ð2Þ

where σC is the yield strength of the collagen layer in tension,JC is the toughness of the collagen layer and EC is its elasticmodulus. Using the experimental values JCE23 kJ m�2,ECE0.5 GPa and σCE60 MPa Eq. (2) gives wE0.8 mm, whichis consistent with experimental observations. Since the spa-cing between the clamps was also in the order of millimeters,it is possible that the inelastic region was constrained by theclamps. Unfortunately it was not possible to increase thespacing between the clamps; these attempts resulted in thereduction of the clamping area and in the sample slipping outof the grips. Constraining this region may have thereforeresulted in underestimating the actual toughness of the scale,an effect similar to cracks propagating in ductile adhesives(Tvergaard and Hutchinson, 1996).

The toughness contribution of the energy dissipated uponloading (ahead of tip) and unloading (behind the tip) is givenby Evans et al. (1986), Barthelat and Rabiei (2011), Gao (2006):

JP ¼ 2wUC ð3Þ

where UC is the density of energy absorbed by the collagenlayer in tension, and w is the half of the width of the inelasticregion. From previous experimental tensile stress–strain datawe measured UCE12 MJ m�3. With this value Eq. (3) predictsJPE19 kJ m�2, which is close to the experimental toughness(JCE23 kJ m�2). The model therefore suggests that the mas-sive inelastic deformations ahead and behind cracks in fishscales are the primary source of their toughness.

7.2. Delamination bridging in the R layer

Another prominent feature observed in the fracture experimentsis the delamination of the collagen fibers within the openingcrack. These delaminated fibers have the potential to generateclosure forces and to bridge the crack walls, in a manner similarto laminated composites (Sorensen et al., 2008) and cortical bone(Nalla et al., 2004). Fig. 8a shows a crack propagating through theradial layer. Experimental observations suggest that a crackpropagating along the radial fibers generate massive delamina-tion of bundles of fibrils from the crack faces. As the crack opensthe delaminated fibrils stretch, generating closure forces on thecrack faces. Because of the very high aspect ratio of the fibers, weneglect the energy stored by shear and bending deformationsand we focus on tensile deformation. Fig. 8b shows an individualfiber (i.e. a bundle of fibrils) which was delaminated from thecrack faces. As the crack opens, the fiber stretches and alsocontinues to delaminate from the crack faces. Fig. 8c shows thesame fiber with the tensile force F exposed. The fiber hasmodulus Ef, and to keep the model simple we assume that thefiber has a square cross section with dimension d � d. Thetoughness of the interface (in energy terms) between fibers isnoted Ji. The mechanics of this system is in many ways similarto the peel test, examined by Kendall (Kendall, 1975) in thesteady state delamination crack propagation regime. By balan-cing the elastic energy released by the fiber with the toughnessof the interface (in energy terms) the Kendall model gives:

Ji ¼12

Fd

� �2 1Ef d

þ Fd

1� cos θð Þ ð4Þ

In our case the extension of the fiber and the delaminationlength are coupled, so the stretch of the fiber can be written as:

ε¼ ll0�1 ð5Þ

The force F carried by the fiber can then be written usingHooke’s law and some trigonometry in Fig. 8c:

u ~1 mm

2/u

l

iJ

dd

fE

dEJ fi /10-4 10-3 10-2 10-1 1 101 102

0l

θ

θ

Fig. 8 – (a) Schematic of a crack propagating along the radial layer (b) fibers delaminated from the crack faces bridge the crack;(c) diagram used to develop the mechanical model; (d) delaminated fiber angle as a function of nondimensional parameterJi/Efd


F¼ d2σf ¼ Ef d2 1

cos θ�1

� �ð6Þ

where σf is the tensile stress in the fiber. Eqs. (4) and (6) cannow be combined, leading to:

JiEf d

¼ 12 cos 2θ

þ cos θ� 32

ð7Þ

This result implies that in the steady state the angle of thedelaminated fiber θ is not a function of crack opening. Theangle depends only on the properties and geometry of the fiber(Ef and d) and on the toughness of the interface (Ji). Fig. 8dillustrate the predictions of Eq. (7). Stiff fibers and weakinterfaces (low Ji/Efd) lead to long delamination distance butsmall delamination angle, with little deformation in the fiber.In contrast, soft fibers bonded by tough interfaces (high Ji/Efd)lead to shorter delamination distance but high delaminationangle, accompanied with high strain in the fiber. We now usethe properties of the collagen fibers and of their interfaces inthe model. Using EfE1 GPa (Shen et al., 2008), dE1 μm fromoptical imaging and JiE40 J m�2 as measured above givesJi/EdE0.04. With this value the model predicts a fiber angleof about 301 within the delaminated region, which is consis-tent with experimental observations (Fig. 5c). With this angle,Eq. (6) predicts a tensile force FE150 μN per fiber, correspond-ing to a stress of σE150 MPa and a strain εE0.15. Independentexperiments on collagen fibrils confirm that collagen fibrilscan undergo such stress and strain without failing (Shen et al.,2008). We now turn our attention to the closure stressgenerated by the delaminated fibers. If the intact fibers are

oriented at a small angle ϕ { θ from the crack line, the densityof fibers crossing the crack face is:

Nf ¼sin ϕ

d2ð8Þ

The closure traction that the fibers exert on the crack faces is:

σb ¼NfF sin θ ð9Þ

Combining Eqs. (6), (8) and (9) gives:

σb ¼ Ef sin ϕ1

cos θ�1

� �sin θ ð10Þ

Finally the effect of the closure stress on toughness can beapproximated with:

Jb � σbu ð11Þwhere u is the crack opening at which the bridging stressvanishes. Using Ef¼1 GPa, ϕE51 and θ¼301 gives a closurestress of σb¼6 MPa. The model therefore suggests that whileexperimental observations show massive delamination of thefibers, the closure stress exerted by the delaminated fibersaccount for a small portion of the tensile strength offish scales. Using a crack opening uE1 mm, Eq. (11) givesJbE6 kJ m�2. Because delamination can exert closure stressover large crack openings and long bridging regions, thecontribution of delamination on toughness is non-negligible,but remains smaller than the process zone tougheningdescribed above. These models are highly idealized, and inreality process zone and delamination and toughening are


probably highly coupled microscale phenomena. Nevertheless,the models provide useful insights into the main mechanismsbehind the high toughness of fish scales.

8. Summary

Teleost fish scales are 500 million year old collagen-basedmaterials constructed to provide flexible armored protection.To fulfill this function fish scales are extremely tough, ahighly desirable properties which also makes fracture testingdifficult. In this work we have shown that individual fishscales from M. saxatilis are indeed notch insensitive. Using anew miniature setup we measured the toughness of thescales along three crack propagation orientations in thewhole scales, and proceeded to measuring the toughness ofthe individual bony and collagen layer which compose thescale. Our results show that these materials are amongst thetoughest biological materials (in energy terms) for whichfracture toughness data is available (Fig. 9). For comparison

Fig. 9 – Toughness-modulus Ashby chart for natural materials. Tscales, bony layer, collagen layer and their interface are shown

we also plotted the toughness of scales from alligator gar. Forthis type of ganoid scale the modulus is about 10 GPa and thetoughness ranges from about 0.8 to 3.6 kJ m�2 (calculatedusing JIC¼KIC

2 /E), depending on the orientation of crack pro-pagation (Yang et al., 2013a,b). Compared to these ganoidscales teleost scales are more compliant because of theirlesser degree of mineralization but they are as tough in termof stress intensity (KC), and one order of magnitude tougher interm of energy (JC). The non-collagenous interfaces whichhold the collagen fibrils together have a much lower (about400 times) toughness. As a result the fibers can detachrelatively easily, which explains the massive delaminationand defibrillation observed in the experiments. Our fracturemodels suggest that inelastic deformations of the collagenfibrils, which operate over regions in the order of 1–2 mmaround the crack tip is the main contributor to toughness, aprocess similar to nacre (Barthelat and Rabiei, 2011) andadvanced engineering polymers (Evans et al., 1986). Delami-nation of the collagen fibers produce a small bridging stressacross the crack faces, but this mechanism can operate over

he properties of mechanical properties for whole teleost fish(adapted from (Wegst and Ashby, 2004)).


large crack openings and we found that its contribution totoughness is not negligible. These findings may suggest newstrategies for light weight protective materials and fabricswhich combine compliance, toughness and resistance topuncture and lacerations. More fundamentally, the propertiesand mechanisms of teleost fish scales provide useful hints onthe behavior of similar collagenous tissues such as lamellarbone, tendon and ligaments.

Acknowledgement

This research was sponsored by a Discovery Grant from theNatural Sciences and Engineering Research Council of Canada.A.K.D was partially supported by a McGill Engineering DoctoralAward (MEDA).

r e f e r e n c e s

ASTM 2003, ASTM D1938–08: Standard Test Method for Tear-Propagation Resistance (Trouser Tear) of Plastic Film and ThinSheeting by a Single-Tear Method.

Barthelat, F., 2007. Biomimetics for next generation materials.Philos. Trans. R. Soc. London, Ser. A 365, 2907–2919.

Barthelat, F., Rabiei, R., 2011. Toughness amplification in naturalcomposites. J. Mech. Phys. Solids 59 (4), 829–840.

Benton, M.J., 2004. Vertebrate Paleontology, third ed. BlackwellScience Ltd..

Bonderer, L.J., Studart, A.R., Gauckler, L.J., 2008. Bioinspireddesign and assembly of platelet reinforced polymer films.Science 319 (5866), 1069–1073.

Browning, A., Ortiz, C., Boyce, M.C., 2013. Mechanics of compositeelasmoid fish scale assemblies and their bioinspiredanalogues. J. Mech. Behav. Biomed. Mater. 19, 75–86.

Bruet, B.J.F., et al., 2008. Materials design principles of ancient fisharmour. Nat. Mater. 7 (9), 748–756.

Chintapalli, R., et al., 2014. Fabrication, testing and modeling of anew flexible armor inspired from natural fish scales andosteoderms. Bioinspiration Biomimetics, 9.

Currey, J.D., 1999. The design of mineralised hard tissues for theirmechanical functions. J. Exp. Biol. 202 (23), 3285–3294.

Currey, J.D., 1999. The design of mineralised hard tissues for theirmechanical functions. J. Exp. Biol. 202 (23), 3285.

Dastjerdi, A.K., et al., 2012. Cohesive behavior of soft biologicaladhesives: experiments and modeling. Acta Biomater. 8 (9),3349–3359.

Dastjerdi, A.K., Rabiei, R., Barthelat, F., 2013. The weak interfaceswithin tough natural composites: experiments on three typesof nacre. J. Mech. Behav. Biomed. Mater. 19, 50–60.

Dastjerdi, A.K., Tan, E., Barthelat, F., 2013. Direct measurement ofthe Cohesive Law of adhesives using a rigid double cantileverbeam technique. Exp. Mech. 53 (9), 1763–1772.

Davis, J.B., Marshall, D.B., Morgan, P.E.D., 2000. Monazite-containing oxide/oxide composites. J. Eur. Ceram. Soc. 20 (5),583–587.

Dimas, L.S., Buehler, M.J., 2012. Influence of geometry onmechanical properties of bio-inspired silica-basedhierarchical materials. Bioinspiration Biomimetics 7 (3).

Evans, A.G., 1997. Design and life prediction issues for high-temperature engineering ceramics and their composites.Acta Mater. 45 (1), 23–40.

Evans, A.G., et al., 1986. Mechanisms of toughening in rubbertoughened polymers. Acta Metall. 34 (1), 79–87.

Evans, A.G., et al., 2001. Model for the robust mechanical behaviorof nacre. J. Mater. Res. 16 (9), 2475–2484.

Fratzl, P., Weinkamer, R., 2007. Nature’s hierarchical materials.

Prog. Mater Sci. 52 (8), 1263–1334.Gao, H.J., 2006. Application of fracture mechanics concepts to

hierarchical biomechanics of bone and bone-like materials.

Int. J. Fract. 138 (1-4), 101–137.Garrano, A.M.C., et al., 2012. On the mechanical behavior of

scales from Cyprinus carpio. J. Mech. Behav. Biomed. Mater.

7, 17–29.Ikoma, T., et al., 2003. Microstructure, mechanical, and

biomimetic properties of fish scales from Pagrus major.

J. Struct. Biol. 142 (3), 327–333.Imbeni, V., et al., 2005. The dentin–enamel junction and the

fracture of human teeth. Nat. Mater. 4 (3), 229–232.Kardong, K.V., 2006. Vertebrates: Comparative Anatomy,

Function, Evolution. McGraw-Hill, Boston.Kendall, K., 1975. Thin-film peeling—elastic term. J. Phys. D: Appl.

Phys. 8 (13), 1449–1452.Ker, R.F., 2007. Mechanics of tendon, from an engineering

perspective. Int. J. Fatigue 29 (6), 1001–1009.Koester, K.J., Ager, J.W., Ritchie, R.O., 2008. The true toughness of

human cortical bone measured with realistically short cracks.

Nat. Mater. 7 (8), 672–677.Lin, Y.S., et al., 2011. Mechanical properties and the laminate

structure of Arapaima gigas scales. J. Mech. Behav. Biomed.

Mater. 4 (7), 1145–1156.Mayer, G., 2005. Rigid biological systems as models for synthetic

composites. Science 310 (5751), 1144–1147.Meyers, M.A., et al., 2008. Biological materials: structure and

mechanical properties. Prog. Mater Sci. 53, 1–206.Meyers, M.A., et al., 2012. Battle in the Amazon: Arapaima versus

Piranha. Adv. Eng. Mater. 14 (5), B279–B288.Mirkhalaf, M., Dastjerdi, A.K., Barthelat, F., 2014. Overcoming the

brittleness of glass through bio-inspiration and micro-

architecture. Nat. Commun., 5.Nalla, R.K., Kinney, J.H., Ritchie, R.O., 2003. Mechanistic fracture

criteria for the failure of human cortical bone. Nat. Mater.

2 (3), 164–168.Nalla, R.K., Kruzic, J.J., Ritchie, R.O., 2004. On the origin of the

toughness of mineralized tissue: microcracking or crack

bridging?. Bone 34 (5), 790–798.Purslow, P.P., 1983. Measurement of the fracture toughness of

extensible connective tissues. J. Mater. Sci. 18 (12),

3591–3598.Shen, Z.L., et al., 2008. Stress–strain experiments on individual

collagen fibrils. Biophys. J. 95 (8), 3956–3963.Sorensen, L., et al., 2008. Bridging tractions in mode I

delamination: measurements and simulations. Compos. Sci.

Technol. 68 (12), 2350–2358.Studart, A.R., 2012. Towards high-performance bioinspired

composites. Adv. Mater. 24 (37), 5024–5044.Thurner, P.J., et al., 2007. High-speed photography of compressed

human trabecular bone correlates whitening to microscopic

damage. Eng. Fract. Mech. 74 (12), 1928–1941.Tvergaard, V., Hutchinson, J.W., 1996. On the toughness of ductile

adhesive joints. J. Mech. Phys. Solids 44 (5), 789–800.Vernerey, F.J., Barthelat, F., 2010. On the mechanics of fishscale

structures. Int. J. Solids Struct. 47 (17), 2268–2275.Weaver, J.C., et al., 2012. The Stomatopod Dactyl Club:

a formidable damage-tolerant biological hammer. Science 336

(6086), 1275–1280.Wegst, U.G.K., Ashby, M.F., 2004. The mechanical efficiency of

natural materials. Philos. Mag. 84 (21), 2167–2186.Yang, W., et al., 2013a. Natural flexible dermal armor. Adv. Mater.

25 (1), 31–48.Yang, W., et al., 2013b. Structure and fracture resistance of

alligator gar (Atractosteus spatula) armored fish scales.

Acta Biomater. 9 (4), 5876–5889.

http://refhub.elsevier.com/S1751-6161(14)00309-9/sbref1















































































































Zhu, D., et al., 2013. Puncture resistance of the scaled skin fromstriped bass: collective mechanisms and inspiration for newflexible armor designs. J. Mech. Behav. Biomed. Mater.24, 30–40.

Zhu, D.J., et al., 2012. Structure and mechanical performance of a“modern” fish scale. Adv. Eng. Mater. 14 (4), B185–B194.

Zimmermann, E.A., et al., 2013. Mechanical adaptability of the

Bouligand-type structure in natural dermal armour.

Nat. Commun., 4.Zok, F.W., 2006. Developments in oxide fiber composites.

J. Am. Ceram. Soc. 89 (11), 3309–3324.












teleost fish scales amongst the toughest collagenous...

Documents